On June 12, 2026, at 5:21pm Eastern, Anthropic received a letter from the U.S. Department of Commerce and, within hours, took its two most capable models — Claude Fable 5 and Claude Mythos 5 — offline for every customer on the planet. The order did not ban the models outright. It barred Anthropic from giving access to "any foreign national," and because Anthropic cannot reliably check nationality on every API call in real time, the only way to comply was to shut the models down for everyone.
This is the first time a U.S. export-control authority has been used to pull a commercially deployed AI model from the market. The core facts are well-corroborated across Anthropic's own statements, Amazon Web Services, and outlets including Fortune, Al Jazeera, Cybersecurity Dive, and the National Law Review. What happens next — whether and how the models come back — is far less settled, and this article keeps the two apart.
What is confirmed
The models, and what they are. Anthropic released both models on June 9, three days before the shutdown. Claude Fable 5 was the public, general-availability model — Anthropic called it "safe for general use" and its most capable model ever made broadly available. Claude Mythos 5 is the same underlying model with safeguards selectively lifted, offered to a small set of vetted cyberdefenders and infrastructure providers. Anthropic describes Mythos 5 as having the strongest cybersecurity capabilities of any model in the world. Fable 5's safety layer routes sensitive requests to Claude Opus 4.8 instead, and Anthropic said this triggered in under 5% of sessions. Both list at $10 per million input tokens and $50 per million output tokens. The order. The directive came from the Commerce Department, attributed in coverage to Commerce Secretary Howard Lutnick, citing national-security export authorities. It prohibits access by any foreign national — and the scope is the part that forced a global shutdown. It applies not only to people outside the United States but to foreign nationals inside it, explicitly including Anthropic's own non-citizen employees. Given that reach, Anthropic concluded it had to disable both models for all users rather than attempt per-user filtering. It also asked AWS to revoke access on Amazon Bedrock, where Fable 5 had been live in the US-East (N. Virginia) and Europe (Stockholm) regions, alongside its availability on the Claude API, Google's Vertex AI, and Microsoft Foundry. The stated reason — and Anthropic's pushback. Officials told Anthropic the government had learned of a technique to bypass Fable 5's safeguards in a way that would unlock Mythos's cybersecurity abilities. Anthropic disputed the severity. In its public statement it described the bypass as "a narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws" — and argued the same prompt could likely elicit comparable behavior from other public models, including OpenAI's GPT-5.5, which face no equivalent restriction. The company called the decision a "misunderstanding" and warned that if a single narrow jailbreak justified recalling a model serving hundreds of millions of people, the same standard would effectively halt new model launches across the entire industry. What still works. Only Fable 5 and Mythos 5 are affected. Anthropic stated plainly that "access to all other Anthropic models will not be affected." That means Claude Opus 4.8, Sonnet 4.6, and Haiku 4.5 remain fully available — including to developers outside the U.S. — as do competing models from OpenAI, Google, and others. If your application was not specifically pinned to Fable 5 or Mythos 5, it kept running.What is not settled: the return
The source most likely to reach you on this story — a wave of "Fable 5 is coming back" coverage — runs ahead of the evidence, so treat the restoration as developing rather than done.
What is genuinely on the record: Anthropic said from the first hour that it was "working to restore access as soon as possible," and its Managing Director of International, Chris Ciauri, told a press conference in Seoul he was "very confident that in the coming days, the models will become available again." That is a statement of intent and optimism, not a confirmed reinstatement.
Beyond that, sources diverge:
The honest summary: the suspension is a hard fact; the restoration is a moving target whose timing, conditions, and geographic scope were still in flux as of this writing. If your work depends on it, watch Anthropic's official status channel, not the headlines.
Why developers should care even if they never used Fable 5
The precedent is the story. A government used export-control law — the same family of authority that has restricted advanced chips from Nvidia and AMD — to reach into a live software service and switch off a specific capability tier. That raises planning questions that outlast this one incident:
The practical takeaway is unglamorous but real: if a single model is load-bearing for your product, you want a fallback path that does not depend on it, and a way to detect availability changes quickly. The teams least disrupted by this episode were the ones already routing across more than one model.
Confirmed vs. inferred — at a glance
Sources
Written by the vybecoding.ai editorial team
Published on June 22, 2026